MaxDuplicateCount = TailDuplicateSize;
if (PreRegAlloc) {
- if (TailBB->empty())
- return false;
- const TargetInstrDesc &TID = TailBB->back().getDesc();
- // Pre-regalloc tail duplication hurts compile time and doesn't help
- // much except for indirect branches and returns.
- if (!TID.isIndirectBranch() && !TID.isReturn())
+ // Pre-regalloc tail duplication hurts compile time and doesn't help
+ // much except for indirect branches.
+ if (TailBB->empty() || !TailBB->back().getDesc().isIndirectBranch())
return false;
// If the target has hardware branch prediction that can handle indirect
// branches, duplicating them can often make them predictable when there
}
// Heuristically, don't tail-duplicate calls if it would expand code size,
// as it's less likely to be worth the extra cost.
- if (InstrCount > 1 && (PreRegAlloc && HasCall))
+ if (InstrCount > 1 && HasCall)
return false;
DEBUG(dbgs() << "\n*** Tail-duplicating BB#" << TailBB->getNumber() << '\n');
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/Support/CFG.h"
-#include "llvm/Support/CommandLine.h"
#include "llvm/Support/ConstantRange.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include <map>
using namespace llvm;
-static cl::opt<bool>
-DupRet("simplifycfg-dup-ret", cl::Hidden, cl::init(false),
- cl::desc("Duplicate return instructions into unconditional branches"));
-
STATISTIC(NumSpeculations, "Number of speculative executed instructions");
namespace {
}
// If we found some, do the transformation!
- if (!UncondBranchPreds.empty() && DupRet) {
+ if (!UncondBranchPreds.empty()) {
while (!UncondBranchPreds.empty()) {
BasicBlock *Pred = UncondBranchPreds.pop_back_val();
DEBUG(dbgs() << "FOLDING: " << *BB
; Same as slightly_more_involved, but block_a is now a CFG diamond with
; fallthrough edges which should be preserved.
-; "callq block_a_merge_func" is tail duped.
; CHECK: yet_more_involved:
; CHECK: jmp .LBB2_1
; CHECK-NEXT: callq bar99
; CHECK-NEXT: callq get
; CHECK-NEXT: cmpl $2999, %eax
-; CHECK-NEXT: jle .LBB2_5
-; CHECK-NEXT: callq block_a_false_func
-; CHECK-NEXT: callq block_a_merge_func
-; CHECK-NEXT: jmp .LBB2_1
-; CHECK-NEXT: .LBB2_5:
+; CHECK-NEXT: jg .LBB2_6
; CHECK-NEXT: callq block_a_true_func
+; CHECK-NEXT: jmp .LBB2_7
+; CHECK-NEXT: .LBB2_6:
+; CHECK-NEXT: callq block_a_false_func
+; CHECK-NEXT: .LBB2_7:
; CHECK-NEXT: callq block_a_merge_func
; CHECK-NEXT: .LBB2_1:
; CHECK-NEXT: callq body
-; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg -S | FileCheck %s
+; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg -S | grep {ret i32 %v1}
+; There should be no uncond branches left.
+; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg -S | not grep {br label}
declare i32 @f1()
declare i32 @f2()
declare void @f3()
define i32 @test(i1 %cond, i1 %cond2, i1 %cond3) {
-; CHECK: test
br i1 %cond, label %T1, label %F1
-; CHECK-NOT: T1:
T1:
%v1 = call i32 @f1()
br label %Merge
br label %Merge
Merge:
-; CHECK: Merge:
-; CHECK: %v1 = call i32 @f1()
-; CHECK-NEXT: %D = and i1 %cond2, %cond3
-; CHECK-NEXT: br i1 %D
%A = phi i1 [true, %T1], [false, %F1]
%B = phi i32 [%v1, %T1], [%v2, %F1]
%C = and i1 %A, %cond2
-; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg -S | FileCheck %s
+; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg -S | grep {ret i32 %v1}
+; There should be no uncond branches left.
+; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg -S | not grep {br label}
declare i32 @f1()
declare i32 @f2()
declare void @f3()
define i32 @test(i1 %cond, i1 %cond2) {
-; CHECK: test
br i1 %cond, label %T1, label %F1
-; CHECK-NOT: T1
T1:
%v1 = call i32 @f1()
br label %Merge
br label %Merge
Merge:
-; CHECK: Merge:
-; CHECK: %v1 = call i32 @f1()
-; CHECK-NEXT: br i1 %cond2
%A = phi i1 [true, %T1], [false, %F1]
%B = phi i32 [%v1, %T1], [%v2, %F1]
%C = and i1 %A, %cond2
-; RUN: opt < %s -jump-threading -S | FileCheck %s
+; RUN: opt < %s -jump-threading -simplifycfg -S | grep {ret i32 1}
; rdar://6402033
; Test that we can thread through the block with the partially redundant load (%2).
target triple = "i386-apple-darwin7"
define i32 @foo(i32* %P) nounwind {
-; CHECK: foo
entry:
%0 = tail call i32 (...)* @f1() nounwind ; <i32> [#uses=1]
%1 = icmp eq i32 %0, 0 ; <i1> [#uses=1]
br i1 %1, label %bb1, label %bb
bb: ; preds = %entry
-; CHECK: bb1.thread:
-; CHECK: store
-; CHECK: br label %bb3
store i32 42, i32* %P, align 4
br label %bb1
ret i32 %res.0
bb3: ; preds = %bb1
-; CHECK: bb3:
-; CHECK: %res.01 = phi i32 [ 1, %bb1.thread ], [ 0, %bb1 ]
-; CHECK: ret i32 %res.01
ret i32 %res.0
}
; CHECK: i64 2, label
; CHECK: i64 3, label
; CHECK: i64 4, label
+; CHECK-NOT: br
; CHECK: }
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
}
+define void @test4() {
+ br label %return
+return:
+ ret void
+; CHECK: @test4
+; CHECK-NEXT: ret void
+}
+@test4g = global i8* blockaddress(@test4, %return)
+
+
; PR5795
define void @test5(i32 %A) {
switch i32 %A, label %return [
; CHECK: i32 16, label %UnifiedReturnBlock
; CHECK: i32 17, label %UnifiedReturnBlock
; CHECK: i32 18, label %UnifiedReturnBlock
-; CHECK: i32 19, label %UnifiedReturnBlock
+; CHECK: i32 19, label %switch.edge
; CHECK: ]
}
; CHECK-NOT: switch
; CHECK: ret void
}
-
-; PR8675
-; rdar://5134905
-define zeroext i1 @test16(i32 %x) nounwind {
-entry:
-; CHECK: @test16
-; CHECK: switch i32 %x, label %lor.rhs [
-; CHECK: i32 1, label %lor.end
-; CHECK: i32 2, label %lor.end
-; CHECK: i32 3, label %lor.end
-; CHECK: ]
- %cmp.i = icmp eq i32 %x, 1
- br i1 %cmp.i, label %lor.end, label %lor.lhs.false
-
-lor.lhs.false:
- %cmp.i2 = icmp eq i32 %x, 2
- br i1 %cmp.i2, label %lor.end, label %lor.rhs
-
-lor.rhs:
- %cmp.i1 = icmp eq i32 %x, 3
- br label %lor.end
-
-lor.end:
- %0 = phi i1 [ true, %lor.lhs.false ], [ true, %entry ], [ %cmp.i1, %lor.rhs ]
- ret i1 %0
-}
-; RUN: opt < %s -simplifycfg -S | FileCheck %s
+; RUN: opt < %s -simplifycfg -S | not grep br
+
%llvm.dbg.anchor.type = type { i32, i32 }
%llvm.dbg.compile_unit.type = type { i32, { }*, i32, i8*, i8*, i8*, i1, i1, i8* }
declare void @llvm.dbg.stoppoint(i32, i32, { }*) nounwind
-define i1 @t({ i32, i32 }* %I) {
-; CHECK: t
-; CHECK: switch i32 %tmp.2.i, label %shortcirc_next.4 [
-; CHECK: i32 14, label %UnifiedReturnBlock
-; CHECK: i32 15, label %UnifiedReturnBlock
-; CHECK: i32 16, label %UnifiedReturnBlock
-; CHECK: i32 17, label %UnifiedReturnBlock
-; CHECK: i32 18, label %UnifiedReturnBlock
-; CHECK: i32 19, label %UnifiedReturnBlock
-; CHECK: ]
+define i1 @_ZN4llvm11SetCondInst7classofEPKNS_11InstructionE({ i32, i32 }* %I) {
entry:
%tmp.1.i = getelementptr { i32, i32 }* %I, i64 0, i32 1 ; <i32*> [#uses=1]
%tmp.2.i = load i32* %tmp.1.i ; <i32> [#uses=6]